Scale Development by Taking into Account Common Method Variance
Cherng George Ding, Ten‐Der Jane, Chiu-Hui Wu, Hang-Rung Lin · Academy of Management Proceedings · 2014
In the present study, an updated procedure for scale development by taking into account common method variance at the item level is proposed, focusing on scale reduction. The CFA marker technique can detect if CMV exists and identify the items contaminated by CMV. We remove the items, one at a time, until none of the items left in the scales for substantive constructs is contaminated. Subsequently, traditional CFA/SEM is conducted, without the threat of CMV, to continue reducing the scale for the focal construct based on the criterion of maximizing convergent validity while maintaining unidimensionality, discriminant validity, and nomological validity. In particular, procedural remedies should be implemented in the stage of item generation to preclude sources of bias and mitigate CMV. Three assumptions for the updated procedure are required. First, the measures are reflective rather than formative. Second, substantive and the marker constructs are uncorrelated and do not interact. Third, a nomological network for substantive constructs is given. Preliminary Monte Carlo simulation has indicated that the CFA marker technique with the method-U model performs well in detecting the items contaminated by CMV. On the other hand, analyses without considering CMV would produce biased parameter estimates, leading to biases in validity assessment. The proposed updated procedure for scale development is recommended to reduce negative impacts of CMV.